28 Prompting Tips That Actually Get Better Results from ChatGPT

WIRED published a roundup of 28 techniques for writing more effective ChatGPT prompts, aimed at everyday users who want noticeably better output without any extra tools. The piece argues that most people treat OpenAI's chatbot like a basic search box, and that deliberate prompt construction, things like specifying a role, requesting a format, or asking the model to reason before answering, produces far more useful results. For business owners using ChatGPT for content, research, or customer communication, the gap between a lazy prompt and a well-built one is measurable.

What happened

WIRED compiled 28 practical prompting techniques for ChatGPT, covering everything from basic phrasing habits to more structured approaches like chain-of-thought reasoning. The article is aimed at regular users, not developers, and requires no API access or paid plugins.

The tips fall into a few broad categories:

• Role assignment. Tell the model to act as a specific expert, a lawyer, a copywriter, a data analyst. This shapes the tone and depth of the response.
• Format control. Ask for bullet points, a table, a numbered list, or a specific word count. ChatGPT defaults to whatever it guesses you want; telling it explicitly saves editing time.
• Step-by-step reasoning. Asking the model to “think through this step by step” before giving an answer reduces confident-sounding errors on logic or calculation tasks.
• Constraints and context. Adding limits (“explain this to a non-technical audience” or “keep it under 150 words”) forces tighter, more targeted output.
• Iterative refinement. Treating the first response as a draft, then asking the model to improve specific parts, rather than starting over, produces better final results faster.

According to WIRED, the underlying principle across all 28 tips is that ChatGPT responds to the quality and specificity of what you put in. Vague questions get vague answers.

Why it matters

ChatGPT has become a standard tool for small business tasks: writing product descriptions, drafting emails, summarising documents, brainstorming ad copy. Most users run a quick query, get a mediocre result, and either accept it or give up. Neither option is great.

Better prompting is essentially free performance improvement. You are not paying for a new plan or a third-party tool. You are just changing how you write the instruction. The payoff shows up in fewer revision cycles, more consistent output, and less time fixing AI-generated text that missed the point.

This also matters for anyone managing staff who use AI tools. If your team is getting poor results and blaming “the AI,” the more likely cause is prompt quality. A short internal guide built around techniques like these could lift output across the board.

Our take

We use ChatGPT daily for client work, and the difference between a bare prompt and a structured one is real. The role-assignment technique alone (“you are a senior SEO copywriter writing for a B2B SaaS audience”) consistently cuts our editing time on first drafts.

That said, there is a ceiling. Prompt engineering does not fix a fundamentally flawed request, and it does not make ChatGPT accurate on topics where the model has gaps. Techniques like step-by-step reasoning reduce hallucinations on logical tasks but do not eliminate them. You still need a human reviewing anything that goes to a client or gets published.

The WIRED list is a solid starting point, but 28 tips is a lot to hold in your head. In practice, four habits cover most of the gains:

1. Assign a role.
2. Specify the output format.
3. Add audience and constraint context.
4. Ask it to reason before concluding on anything complex.

Start there. Add more structure only when a specific task keeps going wrong.

What to do about it

Pick one recurring ChatGPT task your team does this week, a weekly update email, a product description, a meeting summary, and rewrite the prompt using at least three of the techniques above. Compare the output to what you usually get. That one test will tell you more than reading another article about AI productivity.

Source: WIRED · AI